[julienw]

…ner can catch asynchronous work

#97

current main branch
with this patch

I'm especially interested in the SVG version of the benchmark, and looking especially at the PanTheChart steps. Indeed these steps were the most unstable in the SVG version of the benchmark.

With this change I get very stable results in Firefox, and stable results in Chrome.

What do you think @rniwa @flashdesignory @bgrins?

Firefox Before/After

Chrome Before/After

[bgrins]

I haven't checked other browsers but this pretty clearly moves work into React-Stockcharts-SVG.PanTheChart-async in Firefox (with PR: https://share.firefox.dev/3BTd1ni, without PR: https://share.firefox.dev/42YUSAs). And given that we're using this pattern for other workloads it seems like we should do it here, in addition to other updates to make the SVG workload more realistic like in #97.

[camillobruni]

• Maybe thus us the easiest fix for now.
• Ive added an RAF metric in PR BenchmarkRunner test measurement refactoring #28 maybe this would make StockCharts more reliable on it's own?

[julienw]

• Ive added an RAF metric in PR BenchmarkRunner test measurement refactoring #28 maybe this would make StockCharts more reliable on it's own?

I tested with your rAF-based patch. This seems to change things but isn't removing the problem completely:

• When not profiling, I still see the problem in the reported metrics.
• When profiling, the problem doesn't seem to be there anymore.

So there might still be subtle racing problems that I don't fully understand, even with your patch.

[julienw]

I rebased on top of the latest main branch and included the results in Firefox and Chrome in the PR description above.